I have a couple of webservers with a shared filesystem (for /var/lib/php/session and more), identical apache config and enforced https on all connections. We use round robin dns to distribute the load over them. This seems to be working perfectly fine with direct connections, but when accessed trough squid, the pages never fully loads. The browser ends up looping over all elements in the page again and again.. The squid access.log fills up with: 1210543497.362 257 192.168.11.16 TCP_MISS/200 4472 CONNECT rrdnshostname:443 janfrode DIRECT/192.168.41.6 - 1210543497.559 106 192.168.11.16 TCP_MISS/200 3512 CONNECT rrdnshostname:443 janfrode DIRECT/192.168.41.5 - 1210543497.747 174 192.168.11.16 TCP_MISS/200 21485 CONNECT rrdnshostname:443 janfrode DIRECT/192.168.41.6 - 1210543499.581 193 192.168.11.16 TCP_MISS/200 5080 CONNECT rrdnshostname:443 janfrode DIRECT/192.168.41.5 - 1210543500.606 931 192.168.11.16 TCP_MISS/200 4248 CONNECT rrdnshostname:443 janfrode DIRECT/192.168.41.6 - 1210543500.970 235 192.168.11.16 TCP_MISS/200 3880 CONNECT rrdnshostname:443 janfrode DIRECT/192.168.41.5 - 1210543501.342 313 192.168.11.16 TCP_MISS/200 3896 CONNECT rrdnshostname:443 janfrode DIRECT/192.168.41.6 - so I guess this can be a problem with squid trying to distribute the load over both servers, instead of just picking one as the non-squid browser would do. But even without squid, we can see that the browser handles failover to the next host just fine as soon as the dns ttl expires and it picks the second ip-address. Any ideas what might be causing the problem here ? Can we make squid stick with one ip-address for longer periods, instead of going to both addresses simultaneously? -jf
